www.nature.com/scientificreports
OPEN
Received: 21 June 2017 Accepted: 11 September 2017 Published: xx xx xxxx
Ecosystem biomonitoring with eDNA: metabarcoding across the tree of life in a tropical marine environment Michael Stat 1, Megan J. Huggett 2, Rachele Bernasconi2, Joseph D. DiBattista Berry1, Stephen J. Newman1,3, Euan S. Harvey 1 & Michael Bunce1
1
, Tina E.
Effective marine management requires comprehensive data on the status of marine biodiversity. However, efficient methods that can document biodiversity in our oceans are currently lacking. Environmental DNA (eDNA) sourced from seawater offers a new avenue for investigating the biota in marine ecosystems. Here, we investigated the potential of eDNA to inform on the breadth of biodiversity present in a tropical marine environment. Directly sequencing eDNA from seawater using a shotgun approach resulted in only 0.34% of 22.3 million reads assigning to eukaryotes, highlighting the inefficiency of this method for assessing eukaryotic diversity. In contrast, using ‘tree of life’ (ToL) metabarcoding and 20-fold fewer sequencing reads, we could detect 287 families across the major divisions of eukaryotes. Our data also show that the best performing ‘universal’ PCR assay recovered only 44% of the eukaryotes identified across all assays, highlighting the need for multiple metabarcoding assays to catalogue biodiversity. Lastly, focusing on the fish genus Lethrinus, we recovered intra- and inter-specific haplotypes from seawater samples, illustrating that eDNA can be used to explore diversity beyond taxon identifications. Given the sensitivity and low cost of eDNA metabarcoding we advocate this approach be rapidly integrated into biomonitoring programs. Marine ecosystems are under increasing pressure from a variety of anthropogenic stressors, including climate change and fishing activities1–4. Habitat degradation, overexploited fisheries, altered food web dynamics and shifts in community composition highlights the need for effective management of the marine biome in order to preserve and manage ocean resources sustainably into the future4,5. A fundamental tool for providing data to support management of the marine environment is biomonitoring. By targeting indicator species or repeating surveys of specific sites, regular monitoring of the biota allows impacts affecting the abundance and/or alpha and beta diversity to be measured6. However, current approaches to surveying conspicuous faunal elements (e.g. baited remote underwater video, transects, or diver tows) are dependent on suitable field conditions, a particular set of skills, and are limited to the narrow portion of biodiversity recorded. Indeed, community assemblage surveys are often restricted to a single taxonomic group such as fishes or corals, or even a subset of those groups7,8. Considering the growing evidence that total biodiversity promotes healthy ecosystem functions and that sustaining biodiversity represents a practical framework for ecosystem-based management (EBM), there is a compelling need for more comprehensive approaches to monitoring marine biota9–13. Recent advances in genomic technologies afford new opportunities across a broad range of applications in environmental science14–18. One such application lies in the discovery that organisms leave traces of their DNA in the environment, including seawater, and when extracted, is collectively referred to as eDNA19–21. Here, we refer to eDNA as all genetic material that is obtained directly from the environment as in Taberlet et al.21, which depending on sampling methodology, constitutes DNA from whole organisms (i.e. prokaryotes and microscopic 1
Trace and Environmental DNA (TrEnD) Laboratory, Department of Environment and Agriculture, Curtin University, Perth, WA, 6102, Australia. 2Centre for Marine Ecosystems Research, School of Science, Edith Cowan University, Perth, WA, 6027, Australia. 3Western Australian Fisheries and Marine Research Laboratories, Department of Primary Industries and Regional Development, Fisheries Division, Government of Western Australia, PO Box 20, North Beach, WA, 6920, Australia. Correspondence and requests for materials should be addressed to M.S. (email: michael.stat@ curtin.edu.au)
Scientific Reports | 7: 12240 | DOI:10.1038/s41598-017-12501-5
1
www.nature.com/scientificreports/
Figure 1. Assignment of sequences recovered from the shotgun library of eDNA collected from Coral Bay in west Australia. Pie chart segments represent the percentage of sequences that were assigned to taxa using the software MEGAN 5.11.3. Sequences that were assigned to fish were further mined for commonly used DNA barcodes (12S, 16S, 18S, 28S, COI, and cytb); the number of fish barcodes identified in the dataset is displayed in the box insert. eukaryotes), cellular material (i.e. blood, mucous, tissue, faeces, etc.), and that which is released from the cytoplasm as free nucleic acids. The potential utility of eDNA to advance the scientific process is broad, ranging from questions related to species detection, biodiversity assessments, population genetics, reconstruction of past flora and fauna and the detection of invasive marine species22–26. To date however, the capacity of eDNA to inform on ecosystem-wide patterns of biodiversity (i.e. from prokaryotes to higher-order eukaryotes), in a marine environment, remains largely unexplored. A number of methodologies can be employed in the analysis of eDNA. One approach is environmental shotgun sequencing (ESS), which randomly sequences fragmented DNA directly from an environmental sample27. As ESS does not enrich target DNA, the cost associated with sequencing the entire DNA complement present in a sample is prohibitive28. Furthermore, ESS of genetic material recovered from seawater has mostly been applied to the study of prokaryotes29,30 and picoeukaryotes31, and so its utility for characterising eDNA originating from eukaryotes, particularly macroeukaryotes, needs further investigation. To overcome the cost and quantity of DNA that needs to be analysed when using an ESS approach, PCR amplification of target genes (and taxa) on bulk DNA extracts from the environment can be combined with next-generation sequencing (NGS) to provide high-throughput information on the species present, a technique commonly referred to as DNA metabarcoding32. While this approach has proven useful in detecting a high diversity of species from a variety of environmental samples (e.g. soil and water), the influence of PCR-bias on taxonomic recovery and limited correlations to biomass of target species have led researchers to explore the utility of other PCR-free methods33–37. For example, gene-enrichment approaches employ synthetic probes that bind and purify target DNA of interest followed by NGS38. While gene-enrichment is a powerful method in detecting taxa of interest from bulk samples (e.g. synthetic mixtures of macroinvertebrates39), its application across the remaining taxa is currently prohibitive due to the high cost of probes that would capture target genes from all other organisms present in the environment. The capacity of eDNA to inform on eukaryotic diversity from aquatic environments was described in 200820, and yet there is no single study that we are aware of that has explored the utility of eDNA methodologies to assess marine biota at a holistic ecosystem level (i.e. across the tree of life). Accordingly, before temporal and spatial surveys using eDNA can be implemented, the scope and resolution at which metabarcoding can inform on biodiversity in the ocean needs to be validated17. While it has been shown previously that eDNA methodologies are superior to conventional surveying methods in detecting species within aquatic environments through rigorous in silico and in vitro (PCR) analysis of primer sets designed for specific taxa (e.g. amphibians and fish40), the capacity for the analysis of eDNA from seawater to broadly describe total biodiversity in any given sample requires further investigation. Therefore, in this study, we focused on a single tropical coral reef site to explore the ability of eDNA to audit marine biota across the entire tree of life. We chose to sample seawater from Coral Bay in west Australia as it resides within the World Heritage site of Ningaloo Reef, which is renowned for its rich marine biodiversity and enigmatic megafauna, is one of the world’s longest fringing coral reefs, and is therefore of high conservation importance41. Rather than perform rigorous tests on the detection of specific groups of taxa using eDNA34,40,42, we instead focused on assessing the broad potential of eDNA for auditing marine taxa. To appraise the efficacy of different methodologies for the study of eDNA in the ocean, we analysed over 23 million sequences originating from 9 L of filtered seawater and compared the diversity of taxa detected at Coral Bay using ESS and metabarcoding. We further investigated the potential of eDNA to inform on intra-species diversity using mitochondrial haplotype data, with the intention of developing approaches for the use of eDNA in measuring population diversity for a commercially targeted fish genus in Northwestern Australia43.
Results and Discussion
Shotgun sequencing of eDNA from seawater. We built a Nextera XT library from the DNA extracted from 9 L of seawater collected at Coral Bay, generating 22.3 million reads (single end 151 bp fragments) on an Ilumina NextSeq platform using an ESS approach. Of the 22.3 million DNA sequences in the shotgun library that passed quality filtering, only 14.1% could be assigned to taxa using Blastn (Fig. 1), with the highest proportion of hits assigning to bacteria (94.5%), followed by viruses (3.0%), eukaryotes (2.4%) and archaea (0.1%). Of the sequences assigning to eukaryotes, ~5000 reads matched with commonly used DNA barcodes (12S, 16S, 18S, 28S, COI, cytb). The overall low level of taxonomic assignment and the proportionally high number of hits to prokaryotes compared to eukaryotes using the ESS data is not surprising when considering the bias of the existing reference databases and relative abundance of these taxonomic groups in seawater. There are only 3339 eukaryotic, Scientific Reports | 7: 12240 | DOI:10.1038/s41598-017-12501-5
2
www.nature.com/scientificreports/
Figure 2. Taxonomic phylogram of eukaryotic diversity at Coral Bay in west Australia derived from ToLmetabarcoding. Bar graphs indicate the number of families in each phyla characterised at Coral Bay, and are coloured according to kingdom.
72311 prokaryotic and 5646 viral genomes currently on GenBank (accessed 10 August 2016), many of which are sourced from terrestrial organisms. Further, while there is genetic information on more taxa through a growing database of commonly used barcode genes, these regions only represent a small fraction of an organism’s genome. Therefore, considering that the library preparation of eDNA for ESS includes all of the genetic material (i.e. entire genomes) from potentially all of the taxa that were present at Coral Bay, the likelihood of recovering DNA fragments mapping to barcode regions that have previously been sequenced and are publically available is low. The frequency of eukaryotic hits in the ESS data also reflects the relatively low abundance of micro-eukaryotes in our oceans, as there are estimated to be approximately 10000 viruses, 1000 bacteria, and 20 micro-eukaryotes in 1 mL of seawater44. Therefore, due to the higher number of bacteria that are sampled compared to micro-eukaryotes, and the higher number of prokaryotic versus eukaryotic genomes sequenced (72311 to 3339, respectively), the chance of mapping ESS data to bacterial genes is far greater. Finally, the recovery of eDNA from macroscopic eukaryotes (e.g. metazoans) is restricted to DNA contained in biological secretions, larvae or decaying cells, which is in much lower concentration than eDNA derived from prokaryotes and micro-eukaryotes45. To critically evaluate the capacity of the ESS data to inform on the non-microbial fraction of the eDNA library, we focused on recovering fish DNA fragments rather than characterise the total diversity present in the library given their well-described taxonomy relative to other groups. The initial Blastn search using NCBI assigned 1.2% of the eukaryotic sequences to fish (class Actinopterygii, Chondrichthyes and Cyclostomata; ~875 reads of the original 22.3 M reads − 0.00004%), which is a poor relative representation. The majority of these fish reads mapped to genome assemblies, mRNA or phylogenetic informative genes. With regards to assigning taxonomic identity, the most valuable reads are instead those that map on to commonly used reference barcodes46. Searching the 22.3 million reads, only three sequences that are commonly used as DNA barcodes mapped to fish (Fig. 1); two 18S rDNA sequences, which could only resolve to the class Actinopterygii (matching six fishes with equal similarity; Bit Score = 248.348), and one 28S rDNA fragment with similar low-resolving power (Bit Score = 161.786). The low number of reads and lack of resolving power for fish in the ESS dataset showcases the limited ability of ESS data to inform on eukaryotic diversity. Therefore, while shotgun sequencing using NGS platforms may represent the most unbiased way to explore eDNA from seawater, and has proven useful for the study of bacteria47, we demonstrate that the application of ESS for eukaryotes is currently not feasible and lacks resolution compared to metabarcoding (see below). Even for abundant and well-characterised marine taxa such as fish, NGS and ESS was unable to ‘cut through’ the microbial biomass that overwhelms the data that we recovered. While enrichment of target species through capture-probes is possible39, it is currently cost-prohibitive for use in routine monitoring of marine ecosystems due to the unwieldy number of libraries and probes that would be required for coverage across all taxa.
Metabarcoding of eDNA from seawater. The lack of non-microbial taxa detected from the shotgun sequencing data led us to explore the potential of ToL-metabarcoding, which we define as the use of multiple metabarcoding assays to survey a wide array of biotic diversity. Using ten different metabarcoding assays, a total of 1.2 million amplicon reads (that passed quality filtering) were used to generate a multi-taxon eDNA snapshot of the marine biodiversity at Coral Bay, 20-fold less sequencing effort than that used for ESS. Compared to the 14.1% of the ESS data that could be assigned to taxa, 79.7% of the metabarcoding sequence data could be assigned to taxa at Coral Bay (Supplementary Datas 1 and 2). The metabarcoding data was assigned to 434 eukaryotic taxa: 38 phyla, 88 classes, 186 orders and 287 families (Fig. 2; Table 1; Supplementary Data 3). Likewise for prokaryotes, Scientific Reports | 7: 12240 | DOI:10.1038/s41598-017-12501-5
3
www.nature.com/scientificreports/ Kingdom
Phylum
Class
Order
Family
Genus
Annelida
2
7
17
16
12
35
38
Arthropoda
3
Brachiopoda
1
Bryozoa
1
Chaetognatha
Animalia
Fungi
Plantae
Total
1
5
17
45
73
Cnidaria
2
5
9
7
Ctenophore
1
1
1
1
Echinodermata
3
2
3
4
2
1
1 19
5
Mollusca
4
15
22
Nematoda
1
1
1
Nemertea
2
1
2
Platyhelminthes
4
3
2
Porifera
3
6
6
Xenacoelomorpha
1
36
2 6
1
Bigyra
1
2
2
3
Cercozoa
4
4
3
4
Ciliophora
7
15
24
Cryptophyta
3
3
6
Haptophyta
4
5
5
4
Heliozoa
1
1
1
1 36
Myzozoa
3
10
25
Ochrophyta
9
31
27
6
Oomycota
1
3
2
Radiozoa
2
2
1
Ascomycota
3
4
3
Basidiomycota
2
2
1
1
Chytridiomycota
1
1
1
1
Entomophthoromycota
1
1
1
1
Chlorophyta
7
10
11
17
Glaucophyta
1
1
1
Rhodophyta
1
5
7
3
Streptophyta
1
11
11
2
1
1
1
Amoebozoa Protozoa
3 1
8
Chordata
Gastrotricha
Chromista
1 1
Species
Apusozoa
1 2
1
2
3
Choanozoa
2
1
2
3
Picozoa
1
1
1
38
88
186
287
255
54
Table 1. The number of eukaryotic taxa identified within each taxonomic rank at Coral Bay in west Australia via ToL-metabarcoding. 445 Operational Taxonomic Units (OTUs) from 14 phyla, 28 classes, 61 orders and 96 families were detected (Fig. 3; Supplementary Data 3). By using a suite of metabarcoding assays that target different organisms, all the major taxonomic lineages including the Animalia, Fungi, Protozoa, Plantae, Chromista, Bacteria, and Archaea were detected. For example, three classes of vertebrates were recovered from metabarcoding; Actinopterygii (ray-finned fishes), Chondrichthyes (cartilaginous fishes) and Mammalia. The majority of vertebrate diversity was within the class Actinopterygii (41 families), whereas whiptail stingrays (Dasyatidae) and eagle rays (Myliobatidae), as well as dolphins (Delphinidae), made up the families detected in the class Chondrichthyes and Mammalia, respectively. Thirty classes of invertebrates from 16 phyla, including arthropods and benthic organisms such as cnidarians (which includes corals), sponges and bivalves, were also detected from seawater samples. Interestingly, ten classes (31 families) of plants from four phyla were detected, including some of terrestrial origin, which were likely encountered via dispersal of pollen by wind and freshwater48. Micro-eukaryotes, such as fungi and phytoplankton, as well as prokaryotes, including dominant bacterioplankton such as SAR11, SAR 86, SAR116 and OM4349, were also present in the samples. Rather than critically evaluate the efficiency of each PCR assay used, what this study showcases is the capacity to audit the entire spectrum of taxonomic diversity present in a tropical marine environment using DNA samples extracted from seawater. A similar approach using six PCR assays on eDNA extracted from soil samples detected terrestrial prokaryotes and eukaroytes50, and recent studies on eDNA extracted from seawater and settlement plates detected a wide range of eukaryotes51–54. Collectively, Scientific Reports | 7: 12240 | DOI:10.1038/s41598-017-12501-5
4
www.nature.com/scientificreports/
Figure 3. Hierarchical pie chart of prokaryotic diversity at Coral Bay in west Australia derived from ToLmetabarcoding. The inner pie chart represents the relative proportion of families identified in bacteria (blue) and archaea (red). Each segmented circle illustrates the number of taxa (phyla, classes and orders) characterised for both bacteria and archaea, and is scaled according to the number of families within each rank. Dotted lines partition the number of taxa within each phylum, which are named around the circumference of the chart.
these studies indicate that eDNA methodologies can be used on substrates from a variety of environments to assess a broad range of taxa. With further methodological development, ToL-metabarcoding provides compelling evidence for its inclusion into a biomonitoring ‘toolkit’ for marine environments. When comparing ESS to ToL-metabarcoding, it is clear that the latter method is superior in detecting fish taxa using eDNA. While only three barcode sequences could be assigned to the well-characterised Actinopterygii using ESS (and could not resolve to lower taxonomic levels), 69 Actinopterygii taxa were resolved using ToL-metabarcoding, including 33 to the species level. Further, the application of specific primers for the detection of fishes on eDNA collected from water indicates advantages over traditional survey methods. Analysis of eDNA recovered higher numbers of fish taxa in both marine and freshwater systems compared to traditional surveying methods55,56, and when comparing eDNA surveys to known fish diversity present in aquaria (e.g. Okinawa Churaumi Aquarium, 180 fish species), Miya et al.34 were able to identify >90% of taxa from their 12S rRNA sequences. In our data set, we detected small “bait fishes” from the genera Atherinomorus, Engraulis, Hypoatherina, Hyporhamphus, Sardinella, Spratelloides and Strongylura. While these genera are sometimes seen and recorded by other techniques such as Underwater Visual Census (UVC) by SCUBA divers or Baited Remote Underwater Video (BRUV) systems, it is extremely challenging to identify fish to species-level even if the family can be determined visually7. We also detected a number of cryptobenthic fishes with eDNA (e.g. Families Blenniidae, Gobiidae and Pseudochromidae) and fishes that are mostly active at night (e.g. families Apogonidae, Gerreidae and Holocentridae), which are usually not sampled adequately by other techniques7,57,58. Several species from families targeted by recreational and commercial fishers (i.e. Carangidae, Lethrinidae, Lutjanidae, Mugilidae, Epinephelidae) were also recovered from the metabarcoding data. As eDNA methodologies are non-invasive and show high sensitivity and detection capabilities, the application of metabarcoding eDNA for surveying fish in combination with assays that target other taxa are likely to be adopted in future monitoring programs.
The importance and utility of multiple metabarcoding assays. Our data advocates strongly for the use of ToL-metabarcoding as opposed to relying on a single ‘universal’ PCR assay to audit biota in the marine environment. The universal PCR assay based on the (commonly employed) 18S rDNA V4 region detected 191 taxa, which represents only 44% of the total number of taxa characterised from all PCR assays combined (N = 434; Fig. 4). Further, when normalising for sequencing depth, the trend in the number of taxa resolved for each assay at each taxonomic rank is consistent to that depicted in Fig. 4 (Supplementary Data 4), and there is a significant difference in the number of taxa identified between assays at each taxonomic rank (Kruskal-Wallis rank sum test, p
OPEN
Received: 21 June 2017 Accepted: 11 September 2017 Published: xx xx xxxx
Ecosystem biomonitoring with eDNA: metabarcoding across the tree of life in a tropical marine environment Michael Stat 1, Megan J. Huggett 2, Rachele Bernasconi2, Joseph D. DiBattista Berry1, Stephen J. Newman1,3, Euan S. Harvey 1 & Michael Bunce1
1
, Tina E.
Effective marine management requires comprehensive data on the status of marine biodiversity. However, efficient methods that can document biodiversity in our oceans are currently lacking. Environmental DNA (eDNA) sourced from seawater offers a new avenue for investigating the biota in marine ecosystems. Here, we investigated the potential of eDNA to inform on the breadth of biodiversity present in a tropical marine environment. Directly sequencing eDNA from seawater using a shotgun approach resulted in only 0.34% of 22.3 million reads assigning to eukaryotes, highlighting the inefficiency of this method for assessing eukaryotic diversity. In contrast, using ‘tree of life’ (ToL) metabarcoding and 20-fold fewer sequencing reads, we could detect 287 families across the major divisions of eukaryotes. Our data also show that the best performing ‘universal’ PCR assay recovered only 44% of the eukaryotes identified across all assays, highlighting the need for multiple metabarcoding assays to catalogue biodiversity. Lastly, focusing on the fish genus Lethrinus, we recovered intra- and inter-specific haplotypes from seawater samples, illustrating that eDNA can be used to explore diversity beyond taxon identifications. Given the sensitivity and low cost of eDNA metabarcoding we advocate this approach be rapidly integrated into biomonitoring programs. Marine ecosystems are under increasing pressure from a variety of anthropogenic stressors, including climate change and fishing activities1–4. Habitat degradation, overexploited fisheries, altered food web dynamics and shifts in community composition highlights the need for effective management of the marine biome in order to preserve and manage ocean resources sustainably into the future4,5. A fundamental tool for providing data to support management of the marine environment is biomonitoring. By targeting indicator species or repeating surveys of specific sites, regular monitoring of the biota allows impacts affecting the abundance and/or alpha and beta diversity to be measured6. However, current approaches to surveying conspicuous faunal elements (e.g. baited remote underwater video, transects, or diver tows) are dependent on suitable field conditions, a particular set of skills, and are limited to the narrow portion of biodiversity recorded. Indeed, community assemblage surveys are often restricted to a single taxonomic group such as fishes or corals, or even a subset of those groups7,8. Considering the growing evidence that total biodiversity promotes healthy ecosystem functions and that sustaining biodiversity represents a practical framework for ecosystem-based management (EBM), there is a compelling need for more comprehensive approaches to monitoring marine biota9–13. Recent advances in genomic technologies afford new opportunities across a broad range of applications in environmental science14–18. One such application lies in the discovery that organisms leave traces of their DNA in the environment, including seawater, and when extracted, is collectively referred to as eDNA19–21. Here, we refer to eDNA as all genetic material that is obtained directly from the environment as in Taberlet et al.21, which depending on sampling methodology, constitutes DNA from whole organisms (i.e. prokaryotes and microscopic 1
Trace and Environmental DNA (TrEnD) Laboratory, Department of Environment and Agriculture, Curtin University, Perth, WA, 6102, Australia. 2Centre for Marine Ecosystems Research, School of Science, Edith Cowan University, Perth, WA, 6027, Australia. 3Western Australian Fisheries and Marine Research Laboratories, Department of Primary Industries and Regional Development, Fisheries Division, Government of Western Australia, PO Box 20, North Beach, WA, 6920, Australia. Correspondence and requests for materials should be addressed to M.S. (email: michael.stat@ curtin.edu.au)
Scientific Reports | 7: 12240 | DOI:10.1038/s41598-017-12501-5
1
www.nature.com/scientificreports/
Figure 1. Assignment of sequences recovered from the shotgun library of eDNA collected from Coral Bay in west Australia. Pie chart segments represent the percentage of sequences that were assigned to taxa using the software MEGAN 5.11.3. Sequences that were assigned to fish were further mined for commonly used DNA barcodes (12S, 16S, 18S, 28S, COI, and cytb); the number of fish barcodes identified in the dataset is displayed in the box insert. eukaryotes), cellular material (i.e. blood, mucous, tissue, faeces, etc.), and that which is released from the cytoplasm as free nucleic acids. The potential utility of eDNA to advance the scientific process is broad, ranging from questions related to species detection, biodiversity assessments, population genetics, reconstruction of past flora and fauna and the detection of invasive marine species22–26. To date however, the capacity of eDNA to inform on ecosystem-wide patterns of biodiversity (i.e. from prokaryotes to higher-order eukaryotes), in a marine environment, remains largely unexplored. A number of methodologies can be employed in the analysis of eDNA. One approach is environmental shotgun sequencing (ESS), which randomly sequences fragmented DNA directly from an environmental sample27. As ESS does not enrich target DNA, the cost associated with sequencing the entire DNA complement present in a sample is prohibitive28. Furthermore, ESS of genetic material recovered from seawater has mostly been applied to the study of prokaryotes29,30 and picoeukaryotes31, and so its utility for characterising eDNA originating from eukaryotes, particularly macroeukaryotes, needs further investigation. To overcome the cost and quantity of DNA that needs to be analysed when using an ESS approach, PCR amplification of target genes (and taxa) on bulk DNA extracts from the environment can be combined with next-generation sequencing (NGS) to provide high-throughput information on the species present, a technique commonly referred to as DNA metabarcoding32. While this approach has proven useful in detecting a high diversity of species from a variety of environmental samples (e.g. soil and water), the influence of PCR-bias on taxonomic recovery and limited correlations to biomass of target species have led researchers to explore the utility of other PCR-free methods33–37. For example, gene-enrichment approaches employ synthetic probes that bind and purify target DNA of interest followed by NGS38. While gene-enrichment is a powerful method in detecting taxa of interest from bulk samples (e.g. synthetic mixtures of macroinvertebrates39), its application across the remaining taxa is currently prohibitive due to the high cost of probes that would capture target genes from all other organisms present in the environment. The capacity of eDNA to inform on eukaryotic diversity from aquatic environments was described in 200820, and yet there is no single study that we are aware of that has explored the utility of eDNA methodologies to assess marine biota at a holistic ecosystem level (i.e. across the tree of life). Accordingly, before temporal and spatial surveys using eDNA can be implemented, the scope and resolution at which metabarcoding can inform on biodiversity in the ocean needs to be validated17. While it has been shown previously that eDNA methodologies are superior to conventional surveying methods in detecting species within aquatic environments through rigorous in silico and in vitro (PCR) analysis of primer sets designed for specific taxa (e.g. amphibians and fish40), the capacity for the analysis of eDNA from seawater to broadly describe total biodiversity in any given sample requires further investigation. Therefore, in this study, we focused on a single tropical coral reef site to explore the ability of eDNA to audit marine biota across the entire tree of life. We chose to sample seawater from Coral Bay in west Australia as it resides within the World Heritage site of Ningaloo Reef, which is renowned for its rich marine biodiversity and enigmatic megafauna, is one of the world’s longest fringing coral reefs, and is therefore of high conservation importance41. Rather than perform rigorous tests on the detection of specific groups of taxa using eDNA34,40,42, we instead focused on assessing the broad potential of eDNA for auditing marine taxa. To appraise the efficacy of different methodologies for the study of eDNA in the ocean, we analysed over 23 million sequences originating from 9 L of filtered seawater and compared the diversity of taxa detected at Coral Bay using ESS and metabarcoding. We further investigated the potential of eDNA to inform on intra-species diversity using mitochondrial haplotype data, with the intention of developing approaches for the use of eDNA in measuring population diversity for a commercially targeted fish genus in Northwestern Australia43.
Results and Discussion
Shotgun sequencing of eDNA from seawater. We built a Nextera XT library from the DNA extracted from 9 L of seawater collected at Coral Bay, generating 22.3 million reads (single end 151 bp fragments) on an Ilumina NextSeq platform using an ESS approach. Of the 22.3 million DNA sequences in the shotgun library that passed quality filtering, only 14.1% could be assigned to taxa using Blastn (Fig. 1), with the highest proportion of hits assigning to bacteria (94.5%), followed by viruses (3.0%), eukaryotes (2.4%) and archaea (0.1%). Of the sequences assigning to eukaryotes, ~5000 reads matched with commonly used DNA barcodes (12S, 16S, 18S, 28S, COI, cytb). The overall low level of taxonomic assignment and the proportionally high number of hits to prokaryotes compared to eukaryotes using the ESS data is not surprising when considering the bias of the existing reference databases and relative abundance of these taxonomic groups in seawater. There are only 3339 eukaryotic, Scientific Reports | 7: 12240 | DOI:10.1038/s41598-017-12501-5
2
www.nature.com/scientificreports/
Figure 2. Taxonomic phylogram of eukaryotic diversity at Coral Bay in west Australia derived from ToLmetabarcoding. Bar graphs indicate the number of families in each phyla characterised at Coral Bay, and are coloured according to kingdom.
72311 prokaryotic and 5646 viral genomes currently on GenBank (accessed 10 August 2016), many of which are sourced from terrestrial organisms. Further, while there is genetic information on more taxa through a growing database of commonly used barcode genes, these regions only represent a small fraction of an organism’s genome. Therefore, considering that the library preparation of eDNA for ESS includes all of the genetic material (i.e. entire genomes) from potentially all of the taxa that were present at Coral Bay, the likelihood of recovering DNA fragments mapping to barcode regions that have previously been sequenced and are publically available is low. The frequency of eukaryotic hits in the ESS data also reflects the relatively low abundance of micro-eukaryotes in our oceans, as there are estimated to be approximately 10000 viruses, 1000 bacteria, and 20 micro-eukaryotes in 1 mL of seawater44. Therefore, due to the higher number of bacteria that are sampled compared to micro-eukaryotes, and the higher number of prokaryotic versus eukaryotic genomes sequenced (72311 to 3339, respectively), the chance of mapping ESS data to bacterial genes is far greater. Finally, the recovery of eDNA from macroscopic eukaryotes (e.g. metazoans) is restricted to DNA contained in biological secretions, larvae or decaying cells, which is in much lower concentration than eDNA derived from prokaryotes and micro-eukaryotes45. To critically evaluate the capacity of the ESS data to inform on the non-microbial fraction of the eDNA library, we focused on recovering fish DNA fragments rather than characterise the total diversity present in the library given their well-described taxonomy relative to other groups. The initial Blastn search using NCBI assigned 1.2% of the eukaryotic sequences to fish (class Actinopterygii, Chondrichthyes and Cyclostomata; ~875 reads of the original 22.3 M reads − 0.00004%), which is a poor relative representation. The majority of these fish reads mapped to genome assemblies, mRNA or phylogenetic informative genes. With regards to assigning taxonomic identity, the most valuable reads are instead those that map on to commonly used reference barcodes46. Searching the 22.3 million reads, only three sequences that are commonly used as DNA barcodes mapped to fish (Fig. 1); two 18S rDNA sequences, which could only resolve to the class Actinopterygii (matching six fishes with equal similarity; Bit Score = 248.348), and one 28S rDNA fragment with similar low-resolving power (Bit Score = 161.786). The low number of reads and lack of resolving power for fish in the ESS dataset showcases the limited ability of ESS data to inform on eukaryotic diversity. Therefore, while shotgun sequencing using NGS platforms may represent the most unbiased way to explore eDNA from seawater, and has proven useful for the study of bacteria47, we demonstrate that the application of ESS for eukaryotes is currently not feasible and lacks resolution compared to metabarcoding (see below). Even for abundant and well-characterised marine taxa such as fish, NGS and ESS was unable to ‘cut through’ the microbial biomass that overwhelms the data that we recovered. While enrichment of target species through capture-probes is possible39, it is currently cost-prohibitive for use in routine monitoring of marine ecosystems due to the unwieldy number of libraries and probes that would be required for coverage across all taxa.
Metabarcoding of eDNA from seawater. The lack of non-microbial taxa detected from the shotgun sequencing data led us to explore the potential of ToL-metabarcoding, which we define as the use of multiple metabarcoding assays to survey a wide array of biotic diversity. Using ten different metabarcoding assays, a total of 1.2 million amplicon reads (that passed quality filtering) were used to generate a multi-taxon eDNA snapshot of the marine biodiversity at Coral Bay, 20-fold less sequencing effort than that used for ESS. Compared to the 14.1% of the ESS data that could be assigned to taxa, 79.7% of the metabarcoding sequence data could be assigned to taxa at Coral Bay (Supplementary Datas 1 and 2). The metabarcoding data was assigned to 434 eukaryotic taxa: 38 phyla, 88 classes, 186 orders and 287 families (Fig. 2; Table 1; Supplementary Data 3). Likewise for prokaryotes, Scientific Reports | 7: 12240 | DOI:10.1038/s41598-017-12501-5
3
www.nature.com/scientificreports/ Kingdom
Phylum
Class
Order
Family
Genus
Annelida
2
7
17
16
12
35
38
Arthropoda
3
Brachiopoda
1
Bryozoa
1
Chaetognatha
Animalia
Fungi
Plantae
Total
1
5
17
45
73
Cnidaria
2
5
9
7
Ctenophore
1
1
1
1
Echinodermata
3
2
3
4
2
1
1 19
5
Mollusca
4
15
22
Nematoda
1
1
1
Nemertea
2
1
2
Platyhelminthes
4
3
2
Porifera
3
6
6
Xenacoelomorpha
1
36
2 6
1
Bigyra
1
2
2
3
Cercozoa
4
4
3
4
Ciliophora
7
15
24
Cryptophyta
3
3
6
Haptophyta
4
5
5
4
Heliozoa
1
1
1
1 36
Myzozoa
3
10
25
Ochrophyta
9
31
27
6
Oomycota
1
3
2
Radiozoa
2
2
1
Ascomycota
3
4
3
Basidiomycota
2
2
1
1
Chytridiomycota
1
1
1
1
Entomophthoromycota
1
1
1
1
Chlorophyta
7
10
11
17
Glaucophyta
1
1
1
Rhodophyta
1
5
7
3
Streptophyta
1
11
11
2
1
1
1
Amoebozoa Protozoa
3 1
8
Chordata
Gastrotricha
Chromista
1 1
Species
Apusozoa
1 2
1
2
3
Choanozoa
2
1
2
3
Picozoa
1
1
1
38
88
186
287
255
54
Table 1. The number of eukaryotic taxa identified within each taxonomic rank at Coral Bay in west Australia via ToL-metabarcoding. 445 Operational Taxonomic Units (OTUs) from 14 phyla, 28 classes, 61 orders and 96 families were detected (Fig. 3; Supplementary Data 3). By using a suite of metabarcoding assays that target different organisms, all the major taxonomic lineages including the Animalia, Fungi, Protozoa, Plantae, Chromista, Bacteria, and Archaea were detected. For example, three classes of vertebrates were recovered from metabarcoding; Actinopterygii (ray-finned fishes), Chondrichthyes (cartilaginous fishes) and Mammalia. The majority of vertebrate diversity was within the class Actinopterygii (41 families), whereas whiptail stingrays (Dasyatidae) and eagle rays (Myliobatidae), as well as dolphins (Delphinidae), made up the families detected in the class Chondrichthyes and Mammalia, respectively. Thirty classes of invertebrates from 16 phyla, including arthropods and benthic organisms such as cnidarians (which includes corals), sponges and bivalves, were also detected from seawater samples. Interestingly, ten classes (31 families) of plants from four phyla were detected, including some of terrestrial origin, which were likely encountered via dispersal of pollen by wind and freshwater48. Micro-eukaryotes, such as fungi and phytoplankton, as well as prokaryotes, including dominant bacterioplankton such as SAR11, SAR 86, SAR116 and OM4349, were also present in the samples. Rather than critically evaluate the efficiency of each PCR assay used, what this study showcases is the capacity to audit the entire spectrum of taxonomic diversity present in a tropical marine environment using DNA samples extracted from seawater. A similar approach using six PCR assays on eDNA extracted from soil samples detected terrestrial prokaryotes and eukaroytes50, and recent studies on eDNA extracted from seawater and settlement plates detected a wide range of eukaryotes51–54. Collectively, Scientific Reports | 7: 12240 | DOI:10.1038/s41598-017-12501-5
4
www.nature.com/scientificreports/
Figure 3. Hierarchical pie chart of prokaryotic diversity at Coral Bay in west Australia derived from ToLmetabarcoding. The inner pie chart represents the relative proportion of families identified in bacteria (blue) and archaea (red). Each segmented circle illustrates the number of taxa (phyla, classes and orders) characterised for both bacteria and archaea, and is scaled according to the number of families within each rank. Dotted lines partition the number of taxa within each phylum, which are named around the circumference of the chart.
these studies indicate that eDNA methodologies can be used on substrates from a variety of environments to assess a broad range of taxa. With further methodological development, ToL-metabarcoding provides compelling evidence for its inclusion into a biomonitoring ‘toolkit’ for marine environments. When comparing ESS to ToL-metabarcoding, it is clear that the latter method is superior in detecting fish taxa using eDNA. While only three barcode sequences could be assigned to the well-characterised Actinopterygii using ESS (and could not resolve to lower taxonomic levels), 69 Actinopterygii taxa were resolved using ToL-metabarcoding, including 33 to the species level. Further, the application of specific primers for the detection of fishes on eDNA collected from water indicates advantages over traditional survey methods. Analysis of eDNA recovered higher numbers of fish taxa in both marine and freshwater systems compared to traditional surveying methods55,56, and when comparing eDNA surveys to known fish diversity present in aquaria (e.g. Okinawa Churaumi Aquarium, 180 fish species), Miya et al.34 were able to identify >90% of taxa from their 12S rRNA sequences. In our data set, we detected small “bait fishes” from the genera Atherinomorus, Engraulis, Hypoatherina, Hyporhamphus, Sardinella, Spratelloides and Strongylura. While these genera are sometimes seen and recorded by other techniques such as Underwater Visual Census (UVC) by SCUBA divers or Baited Remote Underwater Video (BRUV) systems, it is extremely challenging to identify fish to species-level even if the family can be determined visually7. We also detected a number of cryptobenthic fishes with eDNA (e.g. Families Blenniidae, Gobiidae and Pseudochromidae) and fishes that are mostly active at night (e.g. families Apogonidae, Gerreidae and Holocentridae), which are usually not sampled adequately by other techniques7,57,58. Several species from families targeted by recreational and commercial fishers (i.e. Carangidae, Lethrinidae, Lutjanidae, Mugilidae, Epinephelidae) were also recovered from the metabarcoding data. As eDNA methodologies are non-invasive and show high sensitivity and detection capabilities, the application of metabarcoding eDNA for surveying fish in combination with assays that target other taxa are likely to be adopted in future monitoring programs.
The importance and utility of multiple metabarcoding assays. Our data advocates strongly for the use of ToL-metabarcoding as opposed to relying on a single ‘universal’ PCR assay to audit biota in the marine environment. The universal PCR assay based on the (commonly employed) 18S rDNA V4 region detected 191 taxa, which represents only 44% of the total number of taxa characterised from all PCR assays combined (N = 434; Fig. 4). Further, when normalising for sequencing depth, the trend in the number of taxa resolved for each assay at each taxonomic rank is consistent to that depicted in Fig. 4 (Supplementary Data 4), and there is a significant difference in the number of taxa identified between assays at each taxonomic rank (Kruskal-Wallis rank sum test, p
